Automatically Classifying Animal Behavior

1. A method for analyzing motion of a subject, the method comprising: processing three dimensional video data frames that represent motion of the subject using a computational model to output a first and second set of modules wherein each module in the first set of modules exhibits pose dynamics satisfying a predetermined similarity threshold and comprises 200-900 milliseconds and each module in the second set of modules exhibits pose dynamics satisfying the predetermined similarity threshold and comprises 200-900 milliseconds; and storing, in a memory, the first set of modules referenced to a data identifier that represents a type of behavior.

2. The method of claim 1 , said processing comprises a step of isolating the subject from a background in the three dimensional video data frames.

3. The method of claim 2 , said processing further comprises a step of identifying an orientation of a feature of the subject on the three dimensional video data frames with respect to a coordinate system common to each of the three dimensional video data frames.

4. The method of claim 3 , said processing further comprises a step of modifying the orientation of the feature of the subject in each of the three dimensional video data frames so that the feature is oriented in a same direction with respect to the coordinate system to output a set of aligned frames.

5. The method of claim 4 , said processing further comprises a step of using a principal component analysis (PCA) to output pose dynamics data, wherein the pose dynamics data represents a pose of the subject through principal component space.

6. The method of claim 5 , further comprising a step of displaying a representation of each of the first and second set of the modules that occur with a frequency above a threshold in the three dimensional video data frames.

7. The method of claim 1 , wherein the computational model comprises a vector autoregressive process representing a stereotyped trajectory through PCA space.

8. The method of claim 1 , wherein the computational model comprises modeling transition periods between each module in the first and second set of modules using a Hidden Markov Model.

9. The method of claim 1 , wherein the three dimensional video data frames are processed to output a series of points in a multidimensional vector space, wherein each point represents three dimensional pose dynamics of the subject.

10. The method of claim 1 , wherein the subject is an animal in an animal study.

11. A method for analyzing motion of a subject to separate the motion into modules, the method comprising: pre-processing three dimensional video data that represents the motion of the subject to isolate the subject from a background; identifying an orientation of a feature of the subject on a set of frames of the three dimensional video data with respect to a coordinate system; modifying the orientation of the subject in at least a subset of the set of frames so that the feature is oriented in a same direction with respect to the coordinate system to output a set of aligned frames; processing the set of aligned frames using a principal component analysis to output pose dynamics data, wherein the pose dynamics data represents a pose of the subject through principal component space; processing the set of aligned frames to temporally segment the pose dynamics data into a first and second set of modules wherein each module in the first set of modules exhibits pose dynamics satisfying a predetermined similarity threshold and comprises 200-900 milliseconds and each module in the second set of modules exhibits pose dynamics satisfying the predetermined similarity threshold and comprises 200-900 milliseconds; and displaying a representation of the first and second set of modules.

12. The method of claim 11 , wherein the processing the set of aligned frames step is performed using a model free algorithm.

13. The method of claim 12 , wherein the model free algorithm comprises computing an auto-correlogram.

14. The method of claim 11 , wherein the processing the set of aligned frames is performed using a model based algorithm.

15. The method of claim 14 , wherein the model based algorithm is an auto-regressive hidden Markov model (AR-HMM).

16. A method of classifying a test compound, the method comprising: identifying a test behavioral representation in a test subject after the test compound is administered to the test subject wherein the test behavioral representation comprises a set of three dimensional video data frames partitioned into a first and second set of modules wherein each module in the first set of modules exhibits pose dynamics satisfying a predetermined similarity threshold and comprises 200-900 milliseconds and each module in the second set of modules exhibits pose dynamics satisfying the predetermined similarity threshold and comprises 200-900 millisecond; comparing the test behavioral representation to a reference behavioral representation, associated with a class of drugs; and determining that the test compound belongs to the class of drugs if the test behavioral representation is identified by a classifier as matching the reference behavioral representation.

17. The method of claim 16 , wherein the test behavioral representation is identified by receiving three dimensional video data representing motion of the test subject; processing the three dimensional video data using a computational model to partition the data into at least the first and second set of modules and at least one set of transition periods between each module in the first and second set of modules; and assigning the at least one set of modules to a category that represents a type of behavior.

18. The method of claim 17 , wherein the three dimensional video data is first processed to output a series of points in a multidimensional vector space, wherein each point represents 3D pose dynamics of the test subject.

19. The method of claim 16 , wherein the test compound is selected from the group consisting of a small molecule, an antibody or an antigen-binding fragment thereof, a nucleic acid, a polypeptide, a peptide, a peptidomimetic, a polysaccharide, a monosaccharide, a lipid, a glycosaminoglycan, and a combination thereof.